We previously shared an example of the RAP erroneously and excessively warming surface temperatures during periods of significant wet snow. That event occurred with RAP forecasts in the Raleigh-Durham area on February 16, 2013 – details of that event were shared in a previous blog post: Bizarre Surface/Boundary Layer Forecast by the RAP during the 16 February Snowstorm.
We followed up with some of the model developers about our experience with the RAP and they believe that the problem was related to a snow-radiation bug in the Goddard SW radiation scheme. The shortwave radiation scheme would, in certain circumstances, grossly underdo the amount of attenuation of SW radiation when snow was falling. This would result in an artificial and excessive warming of the surface (think lots of sunshine despite falling snow).The bug is confined to the 3-d snow mixing ratio part of the scheme and does not appear to occur with the 3-d cloud water or 3-d ice segments. A fix for this problem has been identified and is being tested in parallel versions of the RAP at ESRL. It is unclear when the fix will be implemented in the NCEP operational RAP, perhaps this fall or more likely next spring.
This issue may be showing up again this morning across northern Virginia. Moderate to heavy wet snow is falling in the region and is expected to continue throughout the day. While the amount of cold advection at the surface is very limited, the significant precipitation of snow or melting snow should lock in temperatures near or just above freezing. The image below is provided by the Bufkit Warehouse and shows a plot of hourly 2-m temperature forecasts from BUFR guidance for KIAD (Dulles VA) and KCHO (Charlottesville VA) from a variety of models including the NAM, GFS, and RAP. In the plot, the RAP 2-m temperature forecast shown in light green warms to near 40 degrees at both locations this afternoon despite steady significant precipitation. This forecast is remarkably different from all of the other guidance and doesn’t make sense physically with the environment. Additional investigation will be required to explain the source of the odd RAP behavior but the Goddard SW radiation scheme seems like a good place to start.
I want to acknowledge Stan Benjamin and the ESRL RAP development team for their help in identifying this issue and providing details on the RAP and its internal processes. Investigating the issues from the first event and the follow-up with the modelers has provided me with a much greater appreciation of the complexities and potential difficulties with NWP.